Incorporating Feature Selection Methods into Machine Learning-Based Covid-19 Diagnosis

Abstract

The aim of the study is to diagnose Covid-19 by machine learning algorithms using biochemical parameters. In addition to the aim of the study, October selection was performed using 14 different feature selection methods based on the biochemical parameters available to us. As a result of the study, the performance of the algorithms and feature selection methods was evaluated using performance evaluation criteria. The dataset used in the study consists of 100 covid-negative and 121 covid-positive data from a total of 221 patients. The dataset includes 16 biochemical parameters used for the diagnosis of Covid-19. Feature selection methods were used to reduce the number of parameters and perform the classification process. The result of the study shows that the new feature set obtained using feature selection algorithms yields very similar results to the set containing all features. Overall, 5 features obtained from 16 features by feature selection methods yielded the best performance for the K-Nearest Neighbour algorithm with the FSVFS feature selection method of 86.4 %.

Performance evaluation of classifiers for the COVID-19 symptom-based dataset using different feature selection methods

“…However, this study did not explore the use of different classifiers and their impact on accuracy. A different study [27] assessed 14 various feature selection approaches based on biochemical parameters for COVID-19 diagnosis. These approaches encompassed filter methods, spiral methods, and embedded methods.…”